Temporal Workflows

Every significant operation in Steadybase runs as a Temporal workflow — a durable, fault-tolerant execution that survives crashes, retries failures, and maintains state across long-running operations.

Why Temporal?

Traditional API-based AI applications are stateless and fragile:

The 6 Workflows

Drew Coordinator

9-step multi-agent orchestration. The centerpiece workflow that coordinates research, content, and cross-namespace calls.

Client Onboarding

5-step onboarding flow with compliance review and human approval gates.

Lead Qualification

3-step scoring and routing. AI-powered lead scoring (0-100) with intelligent routing.

Content Generation

4-step content creation with memory context. Generates QBR decks, proposals, and call briefs.

Ticket Resolution

AI-powered ticket analysis with SLA tracking and automatic escalation.

Memory Store

Long-running durable KV store. Handles memory reads, writes, compression, and health monitoring.

Workflow Execution Model

Signals and Queries

Temporal workflows communicate through signals (incoming messages that change state) and queries (read-only state inspection):

// Signal: Approve a step in a workflow
await handle.signal('approveStep', { stepId: 'compliance-review' });
 
// Query: Check workflow status without affecting execution
const status = await handle.query('status');

Task Queue

All Steadybase workflows execute on the steadybase-gtm task queue. The Temporal worker process polls this queue and executes workflow and activity code.

Activities

Activities are the individual units of work within a workflow — they're where the actual LLM calls, data fetching, and memory operations happen. Activities can be retried independently without re-executing the entire workflow.

Workflow Configuration

Workflows connect to Temporal Cloud via mTLS:

const connection = await Connection.connect({
  address: process.env.TEMPORAL_ADDRESS,
  tls: {
    clientCertPair: {
      crt: Buffer.from(process.env.TEMPORAL_TLS_CERT, 'base64'),
      key: Buffer.from(process.env.TEMPORAL_TLS_KEY, 'base64'),
    },
  },
});
 
const client = new Client({
  connection,
  namespace: process.env.TEMPORAL_NAMESPACE,
});

See Temporal Cloud Integration for full setup details.